Surface plasmon resonance reflectance imaging technique for near-field (~100 nm) fluidic characterization

Abstract

Surface plasmon resonance (SPR) reflectance imaging technique is devised as a label-free visualization tools to characterize near-field (100 nm) fluidic transport properties. The key idea is that the SPR reflectance intensity varies with the near-field refractive index (RI) of the test fluid, which in turn depends on the micro/nano-fluidic scalar properties, such as concentrations, temperatures, and phases. The SPR sensor techniques have been widely used in many different areas, particularly in the biomedical and biophysical societies. While flow visualization techniques based on RI detection have been extensively well documented (Merzkirch 1987), the use of SPR imaging for fluidic applications has been introduced only recently since the author’s group presented a series of related studies in the past few years. The primary goal of this review article is two-fold: (1) Introduction of the working principles of the SPR imaging as a fluidic sensor, and (2) Presentation of example measurement applications for various fluidic scalar properties using the SPR imaging sensor technique. Section 1 summarizes the history and the basic principle of SPR by focusing on the Kretschmann’s theory and Sect. 2 describes the laboratory SPR imaging system specifically designed for fluidic applications. Section 3 presents the optical and material properties that affect the SPR measurement capabilities and sensitivity. Section 4 presents example applications of the implemented SPR for different near-field characterization problems, including (1) micromixing concentration field, (2) convective/diffusion of salinity distributions, (3) full-field thermometry, and (4) fingerprinting of crystallized nanofluidic self assembly. Sections 5 and 6 discuss the spatial measurement resolutions of the SPR imaging technique and the overall measurement sensitivities, respectively. Section 7 presents a few suggestions to further enhance the SPR measurement accuracy particularly for near-field fluidic characterization.

1 Introduction

This article presents a review of the surface plasmon resonance (SPR) reflectance imaging technique as a new label-free, full-field, and quantitative visualization tool to nonintrusively measure the near-field scalar properties for micro/nano-fluidic applications.

1.1 What is surface plasmon resonance?

At incident angles larger than the critical angle (Fig. 1a), the incident light is completely reflected at the interface, i.e., the total internal reflection (TIR), and no significant amount is refracted into the bulk phase of the rarer medium from the macroscopic point of view. From the microscale point of view, however, the incident light penetrates the interface into the external medium (fluid) and propagates parallel to the surface in the plane of incidence, creating a standing electromagnetic wave field while the reflectance remains unchanged from the incidence. This standing wave field is called an evanescent wave field, which decays substantially within the order of 100 nm from the interface (Hecht 2002).

Fig. 1

Schematic principles of TIR and SPR: a Total internal reflection (TIR); b SPR refection; and c Kreschmann configuration of a three-layered SPR system consisting of the glass prism (1), the metallic gold film (2), and the test fluid medium (3). θ _sp is the SPR angle, d ₂ is the thickness of the gold layer, k _x denotes the wave propagation number of the evanescent wave, and k _sp denotes the wave propagation number of the SPR wave (Kim and Kihm 2007a)

When a thin metal film intervenes between the internal (prism) and external (fluid) mediums (Fig. 1b), the reflectance no longer holds the conservation of intensity. For plane (p)-polarized incident wave field, the created evanescent wave field is absorbed in resonance by free electrons contained in the thin metal film and the plasmon resonance is created and the resulting reflectance is frustrated or reduced from the TIR. The p-polarized evanescent wave is attenuated to trigger coherent fluctuations of free electrons, and this coherent energy conversion of the photons into free electrons is called the surface plasmon phenomenon (Richie 1957).

When the p-polarized wave-vector of the evanescent field wave matches the surface plasmon wave vector at a certain incident angle, called the SPR angle, the excitation of free electrons and amplification of the reradiative emissions will occur at the upper surface of the metal film contacting the fluid medium (Ferrell et al. 1985; Raether 1988). Note that s-polarized wave-vector is not subjected to SPR and is totally reflected.

The intensity of the reflected light is greatly reduced because a portion of the incoming photons are absorbed for the resonant SPR emission. The resulting SPR reflectance R is always less than 1 because of the absorbance, which depends on the optical, material, and contact medium properties of the system. The magnitude of the p-polarized SPR reflectance is known to be extremely sensitive to the refractive index variation of the near-wall fluid region. Indeed, the SPR is known to have the most sensitive RI detection capabilities on the order of 10⁻⁵ RIU (Ho and Lam 2003; Xinglong et al. 2005) and more recently as fine as 10⁻⁸ RIU (Slavik and Homola 2007).

The SPR excitation requires specific conditions for optical properties of the thin metal film in that the real part of its dielectric constant must be negative, and its absolute magnitude must be greater than that of the imaginary part (Johnson and Christy 1972; Lakowicz 2004; Kryukov et al. 1997). There are several noble metals available for SPR applications, including silver, gold, copper, and aluminum. Among them, gold is preferred because of its stability and superior performance in various environmental conditions (Hornauer 1976; Inagaki et al. 1981).

1.2 Penetration depth of SPR wave field

The e ⁻¹-penetration depth of TIR evanescent wave field is given as (Born and Wolf 2003; Hecht 2002; Kihm et al. 2004)

$$ \delta_{\text{EV}} = \frac{\lambda }{{4\pi \sqrt {\left( {n_{1} \sin \theta_{i} } \right)^{2} - n_{3}^{2} } }} $$

(1)

where the subscripts 1 and 3 refer to the prism and the contacting medium (test fluid), respectively (Fig. 1). The penetration depth varies, depending on the related parameters, and it spans less than 100 nm in typical cases of air or water medium.

The penetration depth of SPR into the contact medium is defined as the depth at which the electric field of surface plasmon wave falls to 1/e, which is equivalent to the reciprocal of the absolute value of wavevector amplitude in the z-direction (Peterlinz and Georgiandis 1996; Rothenhausler et al. 1984, Raether 1988), i.e.,

$$ \delta_{\text{SPR}} = \left| {\frac{\lambda }{2\pi }\left( {\frac{{\varepsilon_{2}^{'} + \varepsilon_{3} }}{{\varepsilon_{2}^{'2} }}} \right)^{1/2} } \right| $$

(2)

where the subscripts 2 and 3 refer to the metal film and the contacting medium, respectively, and the superscript “prime” denotes the real part of the dielectric constant of the thin metal film.

When λ = 632.8 nm is incident onto a gold metal film (ε ₂ = −13.2+1.25i), for example, the SPR penetration depth δ _SPR = 192 nm for water medium and 353 nm for air medium. Therefore, the “near-field” characterization herein refers to a fluid layer within an order of a few hundred nanometers measured from the interface. In addition, the illuminating intensity of the SPR wave is about 10 times greater than the ordinary evanescent wave field generated by TIR under the same illumination source strength (Rothenhausler and Knoll 1988; Lakowicz 2004; Libermann and Knoll 2000; Neumann et al. 2002).

1.3 Calculations of SPR reflectance R using Fresnel equation

The SPR reflectance R, which is based on the three-layer configuration of contact medium–metal film–prism substrate (Fig. 1c) is regarded as a function of parameters:

$$ R = R\left( {n_{1} ,\,n_{2} ,\,n_{3} ,\,d_{2} ,\,\lambda_{I} ,\,\theta } \right) $$

(3)

where n is the refractive index, d ₂ is the gold layer thickness, $\lambda_{\text{I}}$ is the incident wave length, and θ is the incident ray angle. The subscripts i = 1, 2, and 3 refer to prism, thin film metal, and test medium, respectively.

For p-polarized incident light, the reflectance R can be predicted from the well-established Fresnel theory based on ray optics (Kurihara and Suzuki 2002; Neff et al. 2006). While the detailed derivations are available from many public archives, the resulting Fresnel equations are given as

$$ R = \left| {r_{p} } \right|^{2} $$

(4)

$$ r_{p} = \frac{{\left[ {r_{12} + r_{23} \exp \left( {2ik_{z2} d_{2} } \right)} \right]}}{{\left[ {1 + r_{12} r_{23} \exp \left( {2ik_{z2} d_{2} } \right)} \right]}} $$

(5)

$$ r_{12} = \frac{{\left( {Z_{2} - Z_{1} } \right)}}{{\left( {Z_{2} + Z_{1} } \right)}},r_{23} = \frac{{Z_{3} - Z_{2} }}{{Z_{3} + Z_{2} }} $$

(6)

$$ Z_{i} = \frac{{\varepsilon_{i} }}{{k_{zi} }} ,\quad k_{zi} = \left[ {\varepsilon_{i} \left( {{\raise0.7ex\hbox{$\omega $} \!\mathord{\left/ {\vphantom {\omega c}}\right.\kern-\nulldelimiterspace} \!\lower0.7ex\hbox{$c$}}} \right)^{2} - k_{x1}^{2} } \right]^{0.5} ,\quad i = 1,2,3 $$

(7)

$$ k_{x1} = {\raise0.7ex\hbox{$\omega $} \!\mathord{\left/ {\vphantom {\omega {c\sin \theta \varepsilon_{1}^{1/2} }}}\right.\kern-\nulldelimiterspace} \!\lower0.7ex\hbox{${c\sin \theta \varepsilon_{1}^{1/2} }$}} $$

(8)

where r _p is the reflection coefficient for the p-polarized incident light (Hecht, 2002), ω is the angular frequency, c is the speed of light in vacuum, and k is the wave number $\left( {k^{2} = k_{x}^{2} + k_{z}^{2} } \right)$ equivalent to the magnitude of the wave vector $\vec{k}$ and ε is the dielectric constant. The dielectric constant of fluid medium is obtained from the relationship ɛ = n ² (Lakowicz 2004).

The SPR reflectance R can be readily determined as a function of the pertinent variables by numerically solving the above Fresnel equations, either by constructing a short MATLAB program or downloading one of user-friendly programs from the website (Cristofolini 2007).

For example, when d ₂ = 47.5-nm, gold layer is illuminated by 632.8-nm waves at 20°C through a BK7 glass (n ₁ = 1.515) prism, calculations carried out by setting R = 0 provide the SPR angle for zero reflectance as 43.5° for air test medium and 70.7° for water test medium (Fig. 2; Table 1). In the case of a 20-nm film thinner than the optimum, less of the evanescent wave is absorbed because of the reduced population of free electrons, and the SPR reflectance is increased (Fig. 2b). In the case of a thicker film (60 nm), the SPR transmission is more attenuated in the z-direction and the resulting reflectance is again increased. When the incident wave length, the SPR angle, the film thickness and RI, and the prism RI are specified, the reflectance R is a function of a sole parameter, RI of the contact medium. Changes of the contact medium RI, in turn, relate to various near-field property changes, including the fluidic concentrations, temperature, salinity, colloidal concentrations, gas–liquid phase changes, to name a few.

Fig. 2

Calculated SPR reflectance R when illuminated by p-polarized incident waves at λ = 632.8-nm through a BK7 glass (n ₁ = 1.515) prism at 20°C: a R for air as test medium for the gold film thickness of 47.5 nm (θ _sp = 43.5°), and b R for water as test medium for different gold film thicknesses of 20, 47.5, and 60 nm (θ _sp = 70.7°)

Table 1 Thermophysical (Lide 2005) and optical properties (Moreels et al. 1984) of tested fluids

1.4 History and uses of SPR

As early as in the beginning of the twentieth century, Wood (1902) described the surface plasma waves as the excitation source for anomalous diffraction on diffraction gratings. In the late 1960s, Otto (1968) and Kretschmann (1971) demonstrated surface plasma waves in silver as excited by attenuated TIR. In the last two decades, remarkable research progress and development activities have been achieved for surface plasmon waves as optical sensors particularly detecting chemical and biological quantities.

Various techniques using the SPR principle are currently applied to research, which is rapidly expanding to comprise a variety of interests, including biochemical, biomedical, pharmaceutical, chemistry, polymer, and various other engineering fields. One of the most active areas of research is in the field of biomedicine, which is exploring protein–protein interactions, substrate–DNA/cell/protein/enzyme joining, receptor–ligand attractions, and a number of other applications of timely importance (Englebienne et al. 2003; Giebel et al. 1999; Lakowicz 2004; Lee et al. 2005a, b; Liu et al. 2006; Nikitin et al. 1999; Salamon et al. 1997; Yuk and Ha 2005; Shumaker-Parry et al. 2004; Zhang et al. 2001).

The most important distinction of SPR is in its label-free detection capability of binding interactions without the need for fluorescence or radioscopic labeling of biomolecules, such as antibody–antigen interactions, protein–protein interactions, target– and probe–DNA interactions, biomolecule–cell receptor interactions, and ligand–receptor interactions (Pathak and Savelkoul 1997). Figure 3 schematically shows one predominantly used configuration of SPR biosensors detecting the reflectance changes associated with binding interactions of biomolecules (Ramanavieius et al. 2005).

Fig. 3

Most widely used configuration of SPR biosensors of prism coupler-based SPR system: 1—target biomolecules in the flow, 2—metal (Au) surface with receptor biomolecules attached, 3—prism, 4—light source, 5—light detector (Ramanavieius et al. 2005)

Other essential applications outside the field of biomedicine include measurements of thin polymer film thicknesses (Knoll 1998; Kotsev et al. 2003; Morgan and Taylor 1994; Nelson et al. 1999), the sensing of specific gas or liquid components (Chadwick and Gal 1993; Lam et al. 2005; Podgorsek and Franke 1998, 2002), point-wise temperature measurements, theoretical modeling for hydrogenated amorphous silicon, silver film, titanium dioxide layer (TiO₂), a mixture of ethanol and glycol (Chadwick and Gal 1993; Chiang et al. 2004; Sharma and Gupta 2006; Zeng et al. 2005), and the nanoscale optics in the near and wide field using localized surface plasmon polarization (Hutter and Fendler 2004; Natan and Lyon 2002; Smolyyaninov 2005; Smolyyaninov et al. 2005a, b).

In addition, diverse techniques are presently being developed and tested using the surface plasmon optical characteristics: Classical SPR (Kretschmann 1971; Raether 1977, 1988; Salamon et al. 1997), SPR Interferometry (Ho and Lam 2003; Xinglong et al. 2005), SP-coupled Emission (Gryczynski et al. 2004; Lakowicz 2004), SP Fluorescence Spectroscopy (Lee et al. 2005a and b; Libermann and Knoll 2000; Liu et al. 2006; Neumann et al. 2002), SPR Microscopy (Berger et al. 1994; Brockman et al. 2000; Rothenhausler and Knoll 1988), Nano-optics of SP polarization (Hutter and Fendler 2004; Natan and Lyon 2002; Smolyyaninov 2005; Smolyyaninov et al. 2005a and b), and nanosized particle characterization (Venkata et al. 2007; Hawes et al. 2007).

Despite the rich development history and wide applications of SPR, its use as a fluidic characterization sensor has been scarce, and only few attempts have been published in using the SPR imaging technology for label-free characterization of near-field fluidic properties.

2 Experimental setup for SPR reflectance imaging

While a variety of SPR detection systems are commercially available worldwide, virtually all of them are designed for very specific purposes, mostly for biomedical applications; none of them allows enough flexibility to be modified for fluidic applications at hand. Therefore, a new SPR configuration is necessary for near-field fluidic characterization.

Figure 4 shows the optical arrangement of the SPR imaging system developed at the University of Tennessee (Kim and Kihm 2006). The set-up uses a white light source that is monochromatized by a narrow bandpass filter in order to minimize the diffraction interference noises (Brockman et al. 2000; Rothenhausler and Knoll 1988). The randomly polarized white light from a 100-watt Tungsten-Halogen lamp collimates onto the p-polarizer, the E-field of which is parallel to the incident plane. The rotating mirror adjusts the incident angle to be optimized for the SPR angle, and the incident ray illuminates onto the 47.5-nm thick gold film deposited on the 2.5-nm thick Titanium adhesion layer that is coated on the 160-μm thick cover glass substrate. The cover glass RI matches the prism RI and the two are interfaced by a thin index-matching oil layer.

Fig. 4

The experimental layout of the SPR imaging system for near-field fluidic characterization using a p-polarized white light source that is narrow-banded at 632.8 nm (Kim and Kihm 2006)

The band pass filter narrows the white light spectra to be centered at 632.8-nm, with its full width half maximum (FWHM) of 10-nm. A state-of-the-art Hamamatsu 14-bit cooled EM-CCD camera (Model C9100-02) uses a long focal distance lens of Mitutoyo M Plan APO 5× (NA = 0.14, f = 40 mm) to record SPR images at an extremely high spatial resolution.

3 Effects of optical properties on SPR reflectance

The most sensitive optical properties that affect SPR reflectance R are the prism RI (n ₁) and the metal film RI (n ₂), while other properties in Eq. 3 result in relatively less substantial effects. The effects of two primary parameters, n ₁ and n ₂ on R, are discussed in Sects. 3.1 and 3.2, respectively.

3.1 RI of the prism materials (n ₁)

The parameter, n ₁, varies widely from the choice of prism materials, such as BK7 (n ₁ = 1.515), LF (1.575), Topaz (1.61), SF10 (1.723), SF11 (1.779), Diamond (2.417), and Iodine Crystal (3.34). Thus, different prism materials can significantly alter the SPR reflectance dependency on the test fluid RI, i.e., R versus n ₃. For a water–glycerin mixture (Fig. 5a; Table 1), as an illustrative example, the reflectance R rapidly increases with increasing glycerin concentration (n ₃) up to 50% in mass before slowing down to reach about 75%, and then slightly decreases for all the seven prism materials (Fig. 5b). Note that when the SPR angle is tuned to water or 0% glycerin for R = 0, the SPR imaging can be useful in measuring the glycerin concentrations up to about 50% concentration.

Fig. 5

SPR reflectance R calculated based on the three-layer Fresnel modeling for the seven different prism materials and aqueous glycerine mixtures contacting a 47.5-nm thick gold film illuminated by p-polarized λ = 632.8 nm: a refractive index variation for water–glycerin mixture, b R versus n ₃ for the SPR angle tuned at 0% glycerine, and c R versus n ₃ for the SPR angle tuned at 100% glycerin

When the SPR angle is tuned to 100% glycerin, alternatively, R shows drastically different dependency on n ₃ (Fig. 5c). For prism materials with RI below 1.605, either the SPR phenomenon does not occur (BK7) or exhibits non-monotonic dependence of R versus n ₃ (LF). The main reason for this ambiguity is because the lateral component of incident ray’s wavevector (k _x ) is smaller than the surface plasmon wavevector’s (k _sp) (Kretschmann 1971). For prism RIs above 1.605, R shows monotonic decreases with increasing glycerin concentration. Therefore, these prisms (Topaz, SF10, SF11, Diamond, and Iodine Crystal) are appropriate to be used for the full dynamic range of the concentration measurements. Care should be taken in selecting a proper prism material as well as the right SPR angle, depending on the measurement purpose.

3.2 RI of the thin metal films (n ₂)

In contrast to the known properties for bulk metals, thin metal film properties are not uniquely defined, as evidenced by the measured dielectric constants in seven different laboratories for basically identical 47.5-nm thick gold films (Fig. 6). The deviations may well occur from different measurement techniques, differences in surface roughness, dimensional uncertainties for the film thickness, and differences in fabrication processes. The largest deviation of ε _2r is approximately 15% (Kolomenskii et al. 1997) from the average while the largest deviation of ε _2i is up to 50% (Peterlinz and Georgiandis 1996) above the average. On the other hand, the SPR optimum angle θ _SPR shows a mere ±2.5% deviation, and this can be partially attributed to the dominating dependence of θ _SPR on ε _2r , with its relatively narrower deviation range compared with ε _2i . R versus T, which is calculated for BK7 using the Kolomenskii’s dielectric constant data, shows the steepest gradients for the entire T range (Kim and Kihm 2007b), and is thus considered most desirable to use for the R–T correlation. Both Snopok et al.'s (2001) and Peterlinz and Georgiandis' (1996) data are considered inappropriate because of the deflections occurring for T > 60°C. The possibly overestimated values of ε _2i for these two cases are believed to cause the undesirable deflections.

Fig. 6

R versus T for seven different dielectric constants that were measured by seven different research groups for thin gold films of approximately 47.5-nm thickness deposited on the top surface of a BK 7 prism (n₂ = 1.515). (Kim and Kihm 2007b)

4 Applications for characterization of micro/nano-fluidic near-field properties

This section presents diverse micro/nano-fluidic applications of using SPR as a near-field characterization scheme. Section 4.1 presents measurements of micromixing concentration field development of ethanol penetrating into water contained in a micro-channel, Sect. 4.2 shows full-field detection of the near-wall salinity profiles for convective/diffusion of a saline droplet in water, Sect. 4.3 presents the SPR-mapped temperature profiles of transient cooling of a hot water droplet in a cold water environment, and Sect. 4.4 discusses the unveiling of hidden complex cavities formed during nanofluidic self assembly.

4.1 Label-free mapping of microfluidic mixing

Micromixing has received great attention in many areas in microscale application of fluid mixing, including biomedical applications (Belda et al. 2005). Measurements of mixing concentration fields almost always relied upon introduction of foreign tracer particles such as fluorescent spheres (Strook et al. 2002) and following them instead of true molecular diffusional mixing. SPR reflectance imaging technique first allowed a label-free detection of microfluidic mixing based on the RI variations with advancement of mixing of two different fluids.

When the SPR angle is fixed to 70.7° (tuned to water, for example), the mixture RI increases with increasing ethanol concentration and the reflectance intensity R also increases up to 80% mass concentration. Thereafter, both mixture RI and R gradually decrease (Kim and Kihm 2006). The solid curve in Fig. 7a demonstrates the calculated SPR intensity increasing with ethanol concentration. The symbols represent measured SPR intensity data and the error bars span the 95% confidence levels of pixel-by-pixel variations for recorded images that are recorded with a 12-bit cooled CCD camera (Hamamatsu Inc, Model C8800-21C). Note that the SPR intensities are presented in normalized pixel gray levels, $\equiv {{\left( {{\text{PGL}} - {\text{PGL}}_{\min } } \right)} \mathord{\left/ {\vphantom {{\left( {{\text{PGL}} - {\text{PGL}}_{\min } } \right)} {\left( {{\text{PGL}}_{\max } - {\text{PGL}}{}_{\min }} \right)}}} \right. \kern-\nulldelimiterspace} {\left( {{\text{PGL}}_{\max } - {\text{PGL}}{}_{\min }} \right)}}$ where PGL_min and PGL_max correspond to pure water and 80% ethanol mixture conditions, respectively. The close agreement of the measurement data with the theory allows for the use of the predicted calibration curve to convert measured SPR image intensities into corresponding ethanol concentration values.

Fig. 7

a The pixel gray level correlation of SPR reflectance R with ethanol mass concentration when BK7 glass prism with 47.5-nm gold metal layer is illuminated by p-polarized white light narrow-banded at 632.8 nm, and b full-field development of near-field ethanol mixture concentrations penetrating into water contained in a microchannel. The left column shows R in pixel gray levels, and the right column shows the corresponding ethanol mass concentration distributions (Kim and Kihm 2007c)

Figure 7b shows the SPR images in pixel gray level contours (the left column) and the contour maps of the corresponding concentration distributions for the timely progress of ethanol penetrating from the left entrance into pure water filled in a micro-channel. The primary drive for ethanol penetration into water is attributed to the capillary phoretic suction (Allen et al. 1998, White 2008) by the small hydraulic radius of the micro-channel entrance (50-μm high × 91-μm wide). The ethanol-water interface is progressively broadened because of the molecular diffusion occurring with the interfacial advancement.

The SPR illumination wave field decays exponentially within a few hundred nanometers into the test field from the gold film surface, which is approximately consistent with the penetration depth of the evanescent wave field (Kihm et al. 2004). Therefore, the SPR images shown in Fig. 7b represent the near-wall concentration fields confined within a thin fluid region of less than 1-μm from the interface. This effective detection of the near-wall concentration field can extend the scope of the study of the viscous sublayer region for turbulent mixing process or the effect of slip boundary condition on mixing.

The aforementioned ambiguity of ethanol mixture RI limits the concentration measurements up to 80%. In the case of glycerin–water mixture, however, the mixture RI increases with increasing glycerin concentration up to 100% (see Fig. 5a). Henceforth, the corresponding SPR intensity monotonically decreases with increasing glycerin concentration when the SPR angle is tuned for 100% glycerin and the prism material RI is set to be larger than 1.6 (Fig. 5c). Therefore, although not presented, the SPR reflectance intensity imaging is available for the whole range from 0 to 100% for a glycerin–water mixture (Table 1).

4.2 Near-field mapping of salinity diffusion

There have been various methods developed for the measurement of salinity. Among these, three stand out for practical considerations: gravimetric, inductive electrical, and conductivity detection methods. The gravimetric method is essentially the direct weighing of the specific gravity of a saline solution; however, this method has the shortcoming of a single-valued representation for a stationary situation. The induced electrical method is fast and considered to be acceptably precise, but for high precision measurement of salinity, the conductivity measurement method is preferred because of its superior measurement stability. The conductivity measurement technique, however, provides only point-wise detection at a time and features invasive electrode probes.

Fu et al. (2004) used the wavelength-tunable SPR microscopy to detect the point-wise changes of refractive index of saline solution, but the technique lacks full-field capability. The SPR microscopic salinity detection tool is implemented and tested by measuring the time-dependent salinity distribution fields in the near-wall region when a small saline droplet (0.8-mm diameter) reaches the bottom of a water pool and spreads by convective-diffusion of salinity.

The solid curve in Fig. 8a shows Fresnel predictions of normalized R for the range of salinity to 10%. The symbols represent measured normalized pixel gray levels showing an average rms deviation of 0.062 from the predictions. The error bars represent the 95% confidence interval for the pixel-by-pixel variations in a single SPR image. Figure 8b presents the dynamic and full-field detection of near-wall salinity when a 0.8-mm diameter aqueous drop containing 10% saline mass falls into a water bath with depth of approximately 3 mm (Kim and Kihm 2007a). The saline drop (SG = 1.02) falls by gravity toward the gold layer surface and undergoes convective (driven by gravity and inertia)-diffusive (driven by viscous and concentration gradients) process into the surrounding water. The last column shows the salinity profiles along the centerline.

Fig. 8

a Experimental correlation of normalized R and salinity, and b full-field and real-time mapping of near-wall salinity profile development when a 0.8-mm diameter saline drop (10% mass) is dropped into a 3-mm deep water reservoir at t = 0 (Kim and Kihm 2007a)

It is well documented that the ring vortex occurs when the droplet liquid is more viscous than the base liquid (Thomson and Newall 1885). The viscosity of the 10% NaCl aqueous solution taken here is approximately 20% higher than that of pure water. The interface between the NaCl aqueous solution and water is broadened because of the molecular diffusion progressively occurring during the convective-diffusion advancement. The evolvement of the saline drops on the gold surface demonstrates capillary-like phenomena (Korteweg’s theory; Joseph 1990) for two miscible fluids of saline and pure water. It is worthwhile noting that the capillary-like phenomena in two miscible fluids are not yet fully understood, and the SPR reflectance imaging system can further delineate the unknown physics of the capillary-like phenomena.

The successfully tested label-free, full-field, and real-time salinity mapping technique overcomes the limitation of single-point detection of practically all the existing salinity detection techniques (Kim and Kihm 2008). The salinity-calibrated SPR microscopy system allows for the detailed and dynamic mapping of the near-wall salinity distributions with a lateral spatial resolution of about 4 μm when a gravity-falling saline drop reaches the bottom of a shallow water pool.

4.3 Near-wall thermometry

Among the many available means of temperature detectors, it is still safe to say that thermocouple (TC) is the most extensive and rigorous temperature probes for wide temperature range (Eckert and Goldstein 1970). The limitations of TC probes, however, are their relatively large spatial resolution, physical intrusiveness in flow, and the point-wise detection of temperature. The laser-induced fluorescence (LIF) technique uses fluorescence dye molecules to map temperature fields, but the presence of dye molecules may influence the characteristics of the test medium hydrodynamically as well as thermally (Kim et al. 2003).

The refractive index values, n ₁, n ₂, and n ₃ in Eq. 3 vary with temperature, and the resulting R is correlated with the contact fluid temperature via n ₃(T). The refractive index of the prism, n ₁, is specified by the prism material, and its temperature dependence is assumed negligibly small (Sharma and Gupta 2006). The temperature dependence of the thin metal (Au) film n ₂(T) is small but not completely negligible.

Limited SPR temperature tests have been conducted only to point-wise detect thin metal film temperature, and very few SPR tests have been done for the case of liquid temperature mapping. Though the temperature dependence of R (Fig. 9a) is extensively documented in many articles, a brief summary of its analysis is repeated here, as this reviewer believes that it provides worthwhile implications of the use of SPR for thermometry in many different applications.

Fig. 9

a Correlation of SPR reflectance intensity R with water temperature for the SPR angle at 90°C, with or without accounting for the temperature dependence of RI of the 47.5-nm thick gold film. Full-field and real-time mapping of transient temperature fields when a hot water droplet (80°C) falls on the cold Au surface (20°C) in b air environment, and c water environment. (Kim and Kihm 2007b)

The dielectric constant of thin metal film is given by Drude model (Chiang et al. 1998; Ozdemir and Turhan-Sayan 2003) as:

$$ \varepsilon \; \equiv \;\left( {n_{r} + in_{i} } \right)^{2} \; = \;\;1 - \frac{{\omega_{\rm p}^{2} }}{{\omega \left( {\omega + i\omega_{\rm c} } \right)}} $$

(9)

where ε, n _r , n _i are the dielectric constant of thin metal film, the real and imaginary parts of the refractive index, respectively. ω is the angular frequency of the illuminating light, and ω _p and ω _c are the plasmon frequency and the collision frequency of thin metal film, respectively. Equation 9 is expanded to provide

$$ n_{r}^{2} \; - \;n_{i}^{2} \; = \;1 - \frac{{\omega_{\text{p}}^{2} }}{{\omega^{2} + \omega_{\text{c}}^{2} }},\;2n_{r} n_{i} = \frac{{\omega {}_{\text{c}}\omega_{\text{p}}^{2} }}{{\omega \left( {\omega^{2} + \omega_{\text{c}}^{2} } \right)}} $$

(10)

The plasma frequency is defined as (Hecht 2002):

$$ \omega_{\text{p}} = \sqrt {\frac{{4\pi N{\text{e}}^{2} }}{{m^{*} }}} $$

(11)

where N and m* are the free electron number density and effective mass of a single electron, respectively. e denotes the electric charge of a single electron. The temperature dependence of ω _p is given as (Chiang et al. 2004):

$$ \omega_{\text{p}} = \omega_{{{\text{p}}0}} \left[ {1 + \gamma_{\text{e}} \left( {T - T_{0} } \right)} \right]^{ - 1/2} $$

(12)

where γ _e is the thermal volume expansion coefficient of thin metal film and T ₀ is the reference temperature (90°C for the present case). It is shown that the temperature dependence of ω _p is negligibly small compared with that of ω _c, and thus, only the latter will be retained in determining the temperature dependence of the thin metal film dielectric constant in Eq. 9.

The collision frequency, ω _c, consists of phonon–electron scattering frequency, ω _cp and electron–electron scattering frequency, ω _ce (Kolomenskii et al. 1997), i.e.,

$$ \begin{aligned} & \omega_{\text{c}} = \omega_{\text{cp}} + \omega_{\text{ce}} \hfill \\ & \omega_{\text{cp}} \left( T \right) = \omega_{0} \left[ {\left( {\frac{2}{5} + 4\left( {\frac{T}{{T_{\text{D}} }}} \right)^{{^{5} }} \int\limits_{0}^{{T_{\text{D}} /T}} {\frac{{z^{4} }}{{e^{z} - 1}}{\text{d}}z} } \right)} \right] \hfill \\ & \omega_{\text{ce}} \left( T \right) = \frac{1}{6}\pi^{4} \frac{\Upgamma \Updelta }{{hE_{\text{F}} }}\left[ {\left( {k_{\text{B}} T} \right)^{{^{2} }} + \left( {\frac{h\omega }{{4\pi^{2} }}} \right)^{{^{2} }} } \right] \hfill \\ \end{aligned} $$

(13)

all the above constants and coefficients are listed in Table 2.

Table 2 Parametric values that are used to calculate the temperature dependent thin metal refractive indices

Using n = n _r  + in _i  = 0.1718 + i3.637 for the 47.5-nm Au thin film at λ = 632.8 nm, the above equations are solved to determine ω _p = 3.7826ω, and ω _c = 0.08802ω where the specified ω = 2πc/λ = 2.9788 × 10¹² rad/s. Substituting these results into Eq. 12 gives ω ₀ = 1.171 × 10¹⁴ rad/s. The SPR curves accounting for the temperature variation effect in Fig. 9a are determined from the aforementioned analysis and calculations (Kihm 2008).

Example results of SPR thermometry measurements are presented for dynamic temperature field developments when a hot water droplet at 80°C contacts the Au film surface at 20°C and spreads, either in an air environment (Fig. 9b) or a water environment (Fig. 9c) (Kim and Kihm 2007b, d). The most favored R–T correlation, based on Kolomanskii’s n ₂, is used to convert the recorded SPR image intensity distributions into corresponding temperature fields. The gradual heat and energy transport under the air environment allows the contact surface shape to remain circular and spread concentrically, and the surface temperature gradually decreases to the environmental level after 16 s. In the cold-water environment, however, the aggressive single-phase mass and energy diffusion of hot water into cold water deforms the contact surface shape and spread, and the contact surface temperature rapidly approaches the environmental level in a relatively short time period of 1 s.

4.4 Unveiling of fingerprints of nanocrystalline self assembly

How nanoparticle crystalline structures assemble themselves has been a largely unknown process, at least for its dynamic and quantitative characterization, while the existence of cavities has been conjectured with partial observation of evidence (Sommer et al. 2004; Sommer 2007; Sommer and Zhu 2007). However, this process is attracting increased attention from researchers because of the important potential applications of these nanostructures, including nanoscale manufacturing and bioprocesses (Deegan et al. 1997; Whitesides and Grzybowski 2002; Bigioni et al. 2006; Rabani et al. 2003; Xu et al. 2007; Hong et al. 2006; Chon et al. 2007; Fen and Stebe 2004)

The existence of hidden complex cavities formed inside a self-assembled nanocrystalline structure is for the first time in real-time being discovered by using SPR near-field refractive index fingerprinting (Kim and Kihm 2009). Furthermore, the fully 3D cavity dimensions have been quantitatively reconstructed by digitally analyzing the R-G-B interference fringes naturally formed by the reflected rays from the cavity inner walls (Fig. 10a).

Fig. 10

a Simultaneous imaging of microscopic dorsal view from the top, near-field SPR fingerprinting, and natural fringe ventral view of the nano-crystallized inner cavity structure, and b schematic representation of a cross-sectional view illustrating the formation of the inner cavity (Kim and Kihm 2009)

Figure 10b shows a cross-sectional view normal to the dashed lines on the SPR images to provide a schematic picture of the self-assembly mechanism. Anchoring of the nanoparticles starts soon after the droplet contacts the substrate surface; contact line is pinned with the radially outward flow (1) when evaporation begins (Motte et al. 2000; Gu et al. 2005; Sommer and Rozzlosnik 2005; Sommer and Pavlath 2007). The progressive evaporation lowers the surface temperature near the top dorsal area due to the latent heat loss, and consequently, a thermocapillary phoretic Marangoni flow (2) is induced toward the dorsal peak region with the higher surface tension (Deegan et al. 2000; Hu and Larson 2006, Shmuylovich et al. 2002). The “replenishing” flow (3) rises from the bottom to conform to a stagnant point (Xu and Luo 2007) and ultimately creates vacuum in the near field to conceive a cavity. The downward forces (4) of the attractive van-der-Waals interactions of nanoparticles with the substrate surface result in multiple inside anchorings and conform to a complex hidden cavity structure.

The surface “tension” action associated with the cavity interface (5) continually expands the cavity areas with progressive evaporation until the nanoparticle concentration exceeds a fluidic limit and surface tension is no longer active. Aquatic evaporation also decreases the interparticular distance and increases the attractive van-der-Waals force (6), overcoming the electrostatic repulsive force between the negatively charged Al₂O₃ nanoparticles (Israelachvili 1992). This enhances the congregation of the nanoparticles and expedites the evaporation. The internally driven thermophoretic flow (7) along the temperature gradients from the ventral cavity ceiling to the dorsal peak area also contributes to further evaporation Maillard et al. 2000).

Figure 11 shows the details of a self-assembled nanocrystalline structure for the slow dryout case of an aqueous droplet containing 10% volume of 47-nm Al₂O₃ nanoparticles. The microscopic dorsal plane image is taken 2 min after the cavity inception (Fig. 11a) when the crystallization is completed. The crystallized glassy surface conforms to a half-toroidal shape with a number of radial crack lines (Duggal et al. 2006; Pauchard and Allain 2003). The crystallized crust (Wang and Evans 2006; Haw et al. 2002) is approximately 2 mm in diameter and 160 μm in maximum height.

Fig. 11

Anatomy of a self-assembled nanocrystalline structure revealing the hidden hollow complex cavities formed when a 2-μl aqueous droplet containing 10% volume of 47-nm diameter Al₂O₃ nanoparticles was allowed to evaporate and crystallize at ambient temperature (21 ± 0.5°C) and humidity (40%) on a gold surface. The microscopic dorsal view (a) shows the crystallized half-toroidal glassy surface of approximately 2.0-mm diameter and 160-μm maximum height. Nonintrusive fingerprints by the near-field SPR imaging (b) clearly evidence the existence of the hidden hollow cavity structures. Note that each cell bounded by crack lines is crystallized to form a single anchoring onto the gold substrate. The destructive image taken with the roof shattered (c) confirms the SPR fingerprints, but lacks the details particularly in the self-pinned edge region. Naturally occurring R-G-B interference fringe map (d) is constructed by incident ray interference when the rays are reflected from both the ventral inner cavity surface and the gold substrate surface, which carry quantitative information on the cavity dimensions. Finally, 3D reconstruction (e) of the R-G-B fringe map (d) unveils the hidden complex cavity structures and completes the detailed 3D topography showing the maximum vertical scale of approximately 0.72 μm while the maximum crest roof thickness reaches 160 μm. The vertical dimension is exaggerated by a factor of 400 compared with the horizontal dimension. (Kim and Kihm 2009)

The SPR image (Fig. 11b) shows a two-dimensional fingerprint of the 3D cavity structure. The dark image areas correspond to crystallized nanoparticles congregated on the aforementioned anchored regions, and the bright background represents the air interface, both inside and outside the crust. The bisected crystallized spots are most likely attributed to the secondary crack that occurred at a later time than the primary cracks. The noncircular forming of the central crystallized spot is due to non-axisymmetric competition of surface tension pulling associated with the multiple and distributed inner cavity growth.

While Fig. 11b shows the non-destructive SPR fingerprint of the cavity structure, Fig. 11c shows an intrusive image of the crystallized spots after the bulky roof layer has been shattered off. This open view concurs with the non-destructive SPR image for the overall locations, shapes, and sizes of the crystallized inner structures, but with noticeably reduced details because of damages and disturbances imposed on the delicate inner structures during the top roof removal. More detrimentally, this intrusive examination did not allow imaging of the detailed structure of the peripheral edge region because the top roof could not be separated without destroying the inner structures.

Figure 11d shows non-destructive R-G-B fringe imaging (at λ = 635 nm, 535 nm, and 465 nm, respectively) of the hidden cavity structures; these fringes are constructed by interference of the rays reflected from the substrate surface (gold) and rays reflected from the ventral cavity inner surface, and they present quantitative information on the 3D details of the cavity. Two neighboring fringes of an identical color represent the locations where the cavity height differential is equal to λ/4. In addition, the spectral sequence of the fringes determines the slope of the inner cavity walls; the slope increases when fringes are seen in the order of …-B-G-R-B-G-R-…, and the slope decreases when fringes are seen as …-B-R-G-B-R-G-… (Kihm and Pratt 1999). Therefore, a digital image processing of the fringes shown in Fig. 11d allows reconstruction of a full 3D layout of the cavity structure. Furthermore, dynamic recording of the fringe field provides comprehensive knowledge of the time-dependent cavity growth.

Figure 11e shows 3D reconstruction of the fringe map of Fig. 11d using elaborate pixel-by-pixel analyses and artificially intelligent slope mapping to unveil, for the first time, the hidden complex cavities formed inside the self-assembled nanocrystalline structures. The maximum gap height corresponding to the cavity ridges is approximately 720 nm, which occupies approximately 0.5% of the crystallized crust total height of 160 μm. Note that the vertical scale of the reconstructed topography is exaggerated by 400 times compared with the horizontal scale in order to present the cavity structure more clearly.

It is notable that the topography of the crystallized cavity structures of nanofluids remarkably resembles the earth formation of mountains and valleys. The formation of the complex inner structure was found to be attributable to multiple cavity inceptions and their competing growth during the aquatic evaporation. This outcome provides a possible way to the better understanding and feasible control of the formation of nanocrystalline inner structures.

5 Spatial measurement resolutions

Because of the plane wave optical components, the lateral resolution of SPR imaging, as in normal wide-field light microscopy, is affected by the well-known diffraction limit length scale, $l_{\min } = {{1.22\lambda f} \mathord{\left/ {\vphantom {{1.22\lambda f} {D\sim \lambda }}} \right. \kern-\nulldelimiterspace} {D\sim \lambda }}$ (Rothenhausler and Knoll 1988), also referred to as a Rayleigh criterion (Hecht 2002). In addition, SP waves occur in a propagating mode and are strongly damped in their direction due to intrinsic dissipation (Inagaki et al. 1981; Rothenhausler et al. 1984) and radiative damping (Raether 1977; Hornauer 1976). The propagation length that a surface plasmon wave travels along the interface, between the test medium and the gold thin film, is defined by decay length L _x that can be regarded as a conservative measure of the spatial measurement resolutions on the imaging plane (Berger et al. 1994; Rothenhausler and Knoll 1988).

The decay length, the distance along the surface where the attenuation of the plasmon field occurs, is proportional to 1/e, the reciprocal of an exponential function, or equivalently, $L_{x} = \frac{1}{{2k^{\prime\prime}_{\text{sp}} }}$, where $k^{\prime\prime}_{\text{sp}}$ is the imaginary part of the surface plasmon wave vector along the surface, i.e., $k_{\text{sp}} = k^{\prime}_{\text{sp}} + ik^{\prime\prime}_{\text{sp}} = k_{\text{o}} \left( {\frac{{\varepsilon_{2} \varepsilon_{3} }}{{\varepsilon_{2} + \varepsilon_{3} }}} \right)^{1/2} = \frac{2\pi }{\lambda }\left( {\frac{{\varepsilon_{2} \varepsilon_{3} }}{{\varepsilon_{2} + \varepsilon_{3} }}} \right)^{1/2}$. The calculated decay length L _x as a function of ethanol concentration is shown in Fig. 12, where the upper and lower limits refer to the elementary uncertainties of λ, ɛ ₂ and ɛ ₃ (see the next section on measurement uncertainties). The nominal experimental parameters are used as ɛ ₂ = −13.2 + i1.25, λ = 632.8 nm, θ = 70.7° (SPR angle for water), and d ₂ = 47.5 nm gold. The lateral resolution is approximately 4.5 μm for the aforementioned case of ethanol concentration measurements and is approximately equivalent to seven pixels of the CCD camera, with each pixel covering about 0.66 μm square area in the physical domain.

Fig. 12

Lateral resolution of SPR imaging as a function of ethanol mass concentration in water based on the SPR decay length L _x . The nominal experimental parameters are used as ɛ ₂ = −13.2 + i1.25, λ = 632.8 nm, θ = 70.7°, and d ₂ = 47.5 nm. The upper and lower limits correspond to the uncertainty ranges of λ, ɛ ₂, and ɛ ₃. (Kim and Kihm 2006)

The line-of-sight resolution of SPR imaging is estimated to be in the order of one nanometer, based on the rapidly decreasing resonance contrast with the z-direction, while the line-of-sight resolution for normal wide-field light microscopy is typically in the order of 10 nm, based on the interference contrast (Berger et al. 1994; Rothenhausler and Knoll 1988). Thus, the SPR technique can provide additional benefits in enhancing the z-resolution; for example, in measuring thin film thicknesses (Knoll 1998; Kotsev et al. 2003; Morgan and Taylor 1994; Nelson et al. 1999) or the separation distance between two interfaces (Giebel et al. 1999; Zhang et al. 2001).

6 Overall measurement uncertainties

The overall measurement uncertainty of SPR reflectance R can be determined based on the single point detection estimation given by Kline and McClintock (1953). Considering R = R (ε ₁, ɛ ₂, ε ₃, d ₂, λ _i , θ, c), where “c” refers to a measured property such as ethanol concentration or salinity, the second-power equation referring to the measurement uncertainty of ω _R is given as:

$$ \omega_{R} = \pm \left[ {\left( {\frac{\partial R}{{\partial \varepsilon_{1} }}\omega_{{\varepsilon_{1} }} } \right)^{2} + \left( {\frac{\partial R}{{\partial \varepsilon_{2} }}\omega_{{\varepsilon_{2} }} } \right)^{2} + \left( {\frac{\partial R}{{\partial \varepsilon_{3} }}\omega_{{\varepsilon_{3} }} } \right)^{2} + \left( {\frac{\partial R}{{\partial \lambda_{i} }}\omega_{{\lambda_{i} }} } \right)^{2} + \left( {\frac{\partial R}{\partial \theta }\omega_{\theta } } \right)^{2} + \left( {\frac{\partial R}{{\partial d_{2} }}\omega_{{d_{2} }} } \right)^{2} } \right]^{{^{1/2} }} $$

(14)

For the nominal experimental parameters of ɛ ₂ = −13.2 + i1.25, λ = 632.8 nm, θ = 70.7°, and d ₂ = 47.5 nm, the elementary uncertainties are given as (Kim and Kihm 2006):

$$ \begin{gathered} \omega_{{\varepsilon_{1} }} = \varepsilon_{1} \times 10^{ - 4} = \pm 2.296 \times 10^{ - 4} ,\;\omega_{{\varepsilon_{2} }} = \pm 0.0438 + \pm i0.148\left ( {\varepsilon_{2} = - 13.2 + i1.25} \right), \hfill \\ \omega_{{\varepsilon_{3} }} = \pm \varepsilon_{3} \times 0.000025\;\left( {\varepsilon_{3} = n_{3}^{2} } \right),\;\;\omega_{{\lambda_{i} }} = \pm 10\;{\text{nm}},\;\omega_{\theta } = \pm 1/60^{\text{o}} = \pm 0.0167^{\text{o}} ,\; \hfill \\ \omega_{{d_{2} }} = 47.5\;{\text{nm}} \times 10\% = \pm 4.75\;{\text{nm}} \hfill \\ \end{gathered} $$

(15)

The variation of the prism dielectric constant $\omega_{{\varepsilon_{1} }}$may be assumed negligible because of its extremely small variations of 10⁻⁴ or 0.01%. The elementary variations of dielectric constant of gold,$\omega_{{\varepsilon_{2} }}$, are ±0.0438 and ±0.148 for its real and imaginary parts, respectively. The concentration of the nominal 100% ethanol has 0.5% variation according to the manufacturer’s specifications. The weighted factor of the variation is considered as the test medium uncertainties, which increase with increasing ethanol concentration. For example, for the test case of 40% ethanol concentration, 0.4*(0.5%) = ±0.2% is estimated for the elementary uncertainty$\omega_{{\varepsilon_{3} }}$. The incident light wavelength variation is adopted from the narrow band pass filter specification with FWHM (Full-width half maximum) of 10 nm. The rotation stage that determines the incident angle has an accuracy of 1/60° = ±0.0167°. The thin film thickness uncertainty $\omega_{{d_{2} }}$is estimated to ±10% as provided by the manufacturer (Playtypus Tech., LLC).

The individual derivatives of R directly using the set of Fresnel equations (Eqs. 3–8) are too complicated to derive in their closed forms, primarily because of the coupling with complex refractive index of gold. Alternatively, numerical estimations for the derivatives can be performed using the first-order finite differential scheme as follows:

$$ \omega_{R} = \pm \left[ {\left( {\mathop {\frac{\Updelta R}{{\Updelta \varepsilon_{2} }}\omega_{{\varepsilon_{2} }} }\limits_{{\mathop :\limits_{{\left( {\text{I}} \right)}} }} } \right)^{2} + \left( {\mathop {\frac{\Updelta R}{{\Updelta \varepsilon_{3} }}\omega_{{\varepsilon_{3} }} }\limits_{{\mathop :\limits_{{\left( {\text{II}} \right)}} }} } \right)^{2} + \left( {\mathop {\frac{\Updelta R}{\Updelta \lambda }\omega_{\lambda } }\limits_{{\mathop :\limits_{{\left( {\text{III}} \right)}} }} } \right)^{2} + \left( {\mathop {\frac{\Updelta R}{\Updelta \theta }\omega_{\theta } }\limits_{{\mathop :\limits_{{\left( {\text{IV}} \right)}} }} } \right)^{2} + \left( {\mathop {\frac{\Updelta R}{{\Updelta d_{2} }}\omega_{{d_{2} }} }\limits_{{\mathop :\limits_{{\left( {\text{V}} \right)}} }} } \right)^{2} } \right]^{1/2} $$

(16)

where each term represents an elementary uncertainty for the dielectric constant of gold thin film (I), the dielectric constant of test sample (II), the incident light wavelength (III), the incident angle (IV), and the thin metal film thickness (V).

The differential ΔR in each elementary term is calculated for the range equivalent to the corresponding elementary uncertainty magnitude with respect to the specified test condition parameter. For example, in the case of term (III), $\Updelta \lambda = \left[ {\lambda + \frac{{\omega_{\lambda } }}{2}} \right] - \left[ {\lambda - \frac{{\omega_{\lambda } }}{2}} \right]$ and $\Updelta R = R\left( {\lambda + \frac{{\omega_{\lambda } }}{2};\varepsilon_{2} ,\varepsilon_{3} ,\theta ,d_{2} } \right) - R\left( {\lambda - \frac{{\omega_{\lambda } }}{2};\varepsilon_{2} ,\varepsilon_{3} ,\theta ,d_{2} } \right)$ where the test parameters are specified as λ = 632.8 nm, and the elementary uncertainty is given as ω _λ  = 10 nm.

Table 3 shows the summary results of different uncertainties for the label-free ethanol mixture concentration measurements, the near-wall salinity field measurements, and the SPR full-field temperature measurements, respectively.

Table 3 Overall measurement uncertainties for SPR reflectance R

${{\omega_{R} } \mathord{\left/ {\vphantom {{\omega_{R} } R}} \right. \kern-\nulldelimiterspace} R}$ for the ethanol concentration measurements can be as high as ±21.5% at 10% ethanol concentrations and as low as ±3.5% when ethanol concentrations are beyond 40%. An important observation is that the uncertainty contributions from the terms (I) and (V) are generally more substantial than the contributions from (III) and (IV), particularly at lower ethanol concentrations. These findings can be considered valuable guideline to design improvement of the SPR system in order to enhance its measurement accuracies.

The salinity measurement uncertainties range from ${{\omega_{R} } \mathord{\left/ {\vphantom {{\omega_{R} } R}} \right. \kern-\nulldelimiterspace} R} = \pm 1. 5 7\%$ at 6% salinity to ±5.26% at 8% salinity. The uncertainty contributions (I and V) originated by the gold films are generally more substantial than the contributions from the rest of the terms in Eq. 16. Therefore, it is observed that the dielectric constant and the thickness of the metal (Au) film are the two most significant factors in determining the overall measurement uncertainties.

The temperature measurement uncertainties are estimated to be ±1.85°C (±3.3% in R) at 20°C, and ±1.41°C (±9.8% in R) at 70°C. The relatively substantial uncertainties larger than ±1.0°C are again attributed, first, to the excessive elementary uncertainties of the dielectric constant of the Au film, and secondly, to the fabrication uncertainties of the Au film thickness.

7 Conclusive remarks

A label-free imaging system has been developed using SPR reflectance that depends on the near-wall refractive index (RI) changes of test medium. The system successfully demonstrates its feasibility as a new tool for full-field mapping of micro/nano-scale scalar properties, including the presented examples of fluidic mixture concentrations, salinity, temperature field, and fingerprints of nanofluidic self-assembly. The author believes that the SPR full-field and label-free imaging technique will contribute to visualization studies of micro- and sub-micro scale fluidics as well as heat transfer phenomena. In addition, the technique carries strong potential as an alternative full-field imaging sensor for diverse bioprocesses in various length- and time-scales.

While the SPR theory has carried rich and long history for many decades, its applications have become popular only for the last decade or so primarily in biomedical, material, and chemistry areas. At present, the SPR technologies are one of the most actively studied optics subjects by physicist, optics scientists, chemists, biomedical imaging scientists as well as materialists for a number of highly promising sensing ideas. The SPR sensing for fluidic applications, however, have not been attempted until the author first published an article in Experiments in Fluids (Kim and Kihm 2006). Thereafter, a series of publications by the author’s group in the past few years have been virtually unique. They are proposing a significant breakthrough for label-free imaging of near-field region of various fluid flows as well.

In order to further enhance the signal-to-noise ratio of SPR imaging and reduce its measurement uncertainties, the followings considerations are recommended for future examination:

• 1. Precise and uniform deposition of metal (Au) layer to minimize the elementary uncertainties associated with non-uniform layer thickness.

• 2. Accurate specification of the dielectric constant of the deposited metal layer.

• 3. Development of prism materials of high RI to enhance the dynamic range and sensitivity of SPR measurements (see table for a few selected examples).

  Prism materials	Refractive index (real)
  Quartz crystal	1.458
  Borosilicate glass (Pyrex)	1.474
  BK7 glass	1.515
  LF glass	1.575
  Topaz	1.61
  SF10 glass	1.723
  SF11 glass	1.779
  Diamond crystal	2.417
  Iodine crystal	3.34

  1. (For visible spectrum range)

• 4. Reduction of the SPR-angle broadening associated with the finite dimension of the imaging field.

• 5. Use of an ultra high-resolution CCD recording system with better than the present 14-bit processor.